Investigating the Improvement of Healthcare Rostering by Training a Neural Network on Personal Preference

Abstract: During recent years, the covid-19 pandemic has shed light on the working conditions for many healthcare workers. These essential people have many aspects in their work life that could be improved, one of which being fair working hours that fit their personal life. By having a schedule that is fit for each employee’s needs and preferences, work-life balance could be improved, and employee turnover reduced. Plain Complex is a Finnish startup company that has developed an AI-based solution for scheduling healthcare personnel in an efficient and fair manner. Their solution takes into account some of the personal preferences of employees, however, it could be improved by taking into account more of the implicit and complex preferences of employees. The aim of this thesis work is to investigate how this could be done using an artificial neural network model (ANN). A feasibility study was conducted, and several ANN models were implemented and tested to find a possible solution. The results showed that it is possible to some extent to find a method for modelling the personal preferences of a person in regard to their schedule. However, the impact that the implemented solution might have on the model created by Plain Complex is limited in its current state. The results showed that the model is able to predict a person’s preferences of a roster to a certain degree, especially if those preferences are similar to what is generally considered to be a good or bad roster.